面向事件的视频语义表示方法  

Research on Video Semantic Representation for Events

作　　者：李旭晖[1] 吴青峰 Li Xuhui;Wu Qingfeng(School of Information Management,Wuhan University,Wuhan 430072)

机构地区：[1]武汉大学信息管理学院,武汉430072

出　　处：《图书情报工作》2020年第10期99-108,共10页Library and Information Service

基　　金：国家自然科学基金重大研究计划"大数据驱动的管理与决策研究"重点支持项目"基于知识关联的金融大数据价值分析、发现及协同创造机制"(项目编号:91646206)研究成果之一。

摘　　要：[目的/意义]视频内容正在影响着我国大量人口的信息生活,视频语义的良好表示是推动当前视频内容研究和视频应用服务向前发展的关键基础。现有的视频语义表示方法存在事件语义表示角度和粒度划分方式单一、缺少灵活的对象语义变化机制的问题,因此探究更有效的视频语义表示方法具有重要意义。[方法/过程]提出面向事件的视频语义表示方法。此方法考虑人的双向认知过程,可以根据不同用户背景和需求从不同角度解读和生成事件语义,并定义相应的语义对象和角色的变化机制。[结果/结论]面向事件的视频语义表示方法具有完整的语义表示框架,支持多角度的事件语义表示,可以灵活地进行属性级、对象级和事件级的语义拓展,能够表示更丰富的视频语义。[Purpose/significance]Video content is affecting the information life of a large number of people in China.The proper representation of video semantic is the key foundation for the current development of video content research and application.The existing methods of semantic representation of video only support the semantic representation of an event from one single perspective and lack the flexible change mechanism of relevant semantic objects,which results in insufficient semantic representation.So it is important to explore more effective video semantic representation methods.[Method/process]This paper proposed a video semantic representation method for events.This method considered the bidirectional nature of human cognitive processes and adopted a scalable way to support multi-perspective interpretation of event semantic.A change mechanism of number and semantic is designed to support relevant objects included in events.[Result/conclusion]This method has a complete semantic representation framework,which can effectively support multi-perspective interpretation of video events.It flexibly supports attribute-level,object-level,and event-level semantic extensions.Generally it can represent richer video semantics than existing methods.

关 键 词：视频语义表示 多角度 语义拓展 

分 类 号：G250[文化科学—图书馆学]